Business Problem

In today’s competitive market landscape, understanding the effectiveness of a sales team’s efforts is paramount to sustain and grow a successful business. Although businesses may experience consistent sales growth, it has become challenging to pinpoint the specific factors that drive success and those that need optimization. Current approach to measuring sales team performance relies primarily on historical sales data and subjective evaluations. However, this approach lacks the necessary granularity to comprehend the actual impact of individual sales team’s actions and strategies on overall sales outcomes.

To overcome these limitations, the sales team needs a data-driven model that can accurately predict the sales generated by them based on their various activities, including prospecting, lead follow-ups, product demonstrations, negotiation strategies, and customer relationship management. The model will provide a comprehensive understanding of which sales tactics yield the highest conversion rates, lead to larger deal sizes, and ultimately contribute to the most significant revenue generation.

The implementation of this predictive model will empower the sales team by providing actionable insights and enabling evidence-based decision-making. By quantifying the relationship between their efforts and sales outcomes, the sales team can focus on the most effective strategies, allocate resources efficiently, and tailor their approaches based on customer preferences and market trends. Additionally, the model will help identify underperforming areas, allowing for targeted training and coaching, thereby enhancing the overall performance and productivity of the sales team.

By leveraging data-driven intelligence, businesses can remain competitive in the market, capitalize on emerging opportunities, and ensure the long-term success of the sales team and the company as a whole.

Business Objective

The primary business objective is to develop and implement a data-driven predictive model that accurately assesses the impact of the sales team’s efforts on generating sales. The model will enable the sales team to gain actionable insights, optimize their strategies, and make evidence-based decisions to improve overall sales performance and productivity.

Data Understanding & Exploration

This project aims to improve the productivity of sales representatives. The dataset contains crucial information on various sales performance metrics and customer engagement statistics for an FMCG company spanning 4years. Our objective is to identify key insights, patterns, and actionable strategies that will enable sales reps to achieve higher sales efficiency, customer reach, and conversion rates.

The dataset contains the following variables:

  • Rep_Id: A unique identifier assigned to each sales representative.
  • Gender: The gender of the sales representative.
  • Location: The geographical location where the sales representative operates.
  • Sales: The total sales achieved by each sales representative.
  • Reach: The proportion of customers reached compared to the total market potential.
  • Conv_Rate: The rate at which calls made by the sales representative lead to successful conversions.
  • Growth: Previous growth performance of the representative, considering the sales achieved over the past three months.
  • Repeat_Cust: The proportion of customers who made purchases in both the previous and current months.
  • Repeat_Sales: The proportion of total sales generated by repeat customers.
  • Visit_Freq: The average number of customer visits made by the sales representative per month.
  • Workload: The market potential or demand of the areas assigned to each sales representative.
  • ContactTime: The average time spent engaging with customers on a monthly basis.
  • PreVisit: The number of customers who were visited both in the current month and the previous week but did not make a purchase.
  • Basket_Value: The average sales amount obtained from each transaction.
  • Product_Depth: The average number of products purchased by each customer in a month.
  • Avgcalls_Bill: The average number of customer visits made before a purchase occurs.
  • Avgtime_Calls: The average time gap in days between each customer call before a purchase takes place.

These variables collectively provide a comprehensive view of the sales representatives’ activities, customer interactions, and performance metrics. Analyzing these variables can help identify patterns, trends, and opportunities for improving sales productivity.

Figure 1: Data Profile Summary

 

Exploratory Data Analysis

In order to systematically investigate our predictor variables, I classified them into two distinct groupings: metrics pertaining to the effectiveness of the sales team, and metrics related to the efficiency of the sales team. This categorization will enable a more structured and insightful exploration of the data, allowing us to gain a comprehensive understanding of both the team’s performance in driving sales and their proficiency in utilizing resources.

Given our variables are mostly numeric, a critical facet involves the examination of QQ plots for each variable within the efficiency metrics. QQ plots offer valuable insights into the distribution characteristics of the variables under consideration. Upon closer examination, it is evident that these variables do not conform to a normal distribution. Specifically, we observe a noteworthy phenomenon wherein a greater number of data points exhibit deviations from the expected line, particularly at the tails of the distribution.
This departure from normality underscores the need for specialized statistical approaches that take into account the non-normal nature of the data. By acknowledging this inherent distributional behavior, we can subsequently tailor our analytical strategies and interventions to better align with the observed data patterns, ultimately leading to more accurate and effective enhancements in sales team productivity.

 

 

On closer inspection, the variables’ distributions diverge from the normal distribution. Notably, a substantial proportion of data points display deviations from the expected line, particularly at the distribution’s tails. This departure from normality highlights the necessity for specialized statistical methods capable of accommodating the non-normal data distribution.

 

 

 

Correlation Analysis

In this project, I performed a comprehensive correlation analysis aimed at extracting meaningful insights into the drivers impacting the effectiveness of sales representatives. Through the utilization of correlation funnel charts and predictive power scores, we successfully pinpointed noteworthy connections between different variables and the overall performance of our sales representatives. Notably, the correlation funnel chart and the predictive power score chart exhibited a high degree of consistency, affirming the significance of variables such as Product Depth, Basket Size, Visit Frequency, and Conversion Rate. These findings collectively underscored the critical role played by these factors in influencing the productivity of our sales representatives.

Model Building & Diagnotics

In my model building process, I employ the tidymodels suite of libraries, which provides a structured framework for robust analysis. Initially, I partition my dataset into training and testing sets, a crucial step to ensure the evaluation of model generalizability. To enhance the model’s performance, I create a 10-fold cross-validation dataset from the training set, facilitating the optimization of hyperparameters.

During the data preprocessing phase, I apply a logarithmic transformation to the numeric features via the tidymodels’ recipe functionality. This transformation often leads to improved model performance by addressing skewed data distributions.

For the core modeling, I leverage the random forest algorithm, a powerful ensemble technique. To systematically explore the hyperparameter space and identify the most optimal configuration, I conducted an efficient grid search using a racing with ANOVA models strategy. This approach allows for a comprehensive assessment of various parameter combinations, ultimately leading to the selection of the best-performing model.

Model Performance on Dataset
.metric .estimator Train Dataset Test Dataset
rsq Random Forest 0.955 0.803
mae Random Forest 0.298 0.614
rmse Random Forest 0.574 1.298
rpiq Random Forest 1.293 1.293

From the prediction-residual plot above, it is obvious our model could not capture all the trend in the data especially at the higher data point space. This could explain why our model performance dropped from an rsquare of 95% in our train dataset to 80% in the test dataset.

Model Explaination

Variable Importance

Figure 2: Prediction Contribution

The feature importance chart highlights the key drivers impacting sales productivity improvement. Basket size holds the highest significance, indicating that larger purchase quantities positively influence sales. Following closely is visit frequency, emphasizing the importance of engaging customers more frequently. Product depth ranks third, suggesting that a diverse product range positively contributes to sales. The average number of calls made before purchase comes next, underlining the value of building relationships through multiple interactions. Conversion rate, as expected, plays a crucial role, demonstrating the effectiveness of turning leads into actual sales.

Workload and average time between calls follow, showcasing the delicate balance between efficiency and maintaining customer connections. The proportion of repeat sales over total sales emphasizes the need for customer retention. Reach and contact time also have their place, reflecting the significance of expanding outreach and maximizing customer interaction duration. Lastly, the impact of location indicates potential variations in sales based on geographical factors. These insights collectively guide strategies to enhance sales productivity.

Predictor-Response Relationship

In today’s dynamic business landscape where competition is fierce and market conditions are ever-evolving, optimizing sales team productivity has become a critical factor for an organisation’s success. Partial Dependence Plots are a powerful visualization tool that enables us to interpret the effect of a specific feature while controlling for the influence of other variables. In this project we would generate partial dependence plots for each feature to uncover valuable insights into the characteristics and behaviours that contribute most significantly to sales team productivity.

Our comprehensive analysis delves into the relationship between sales representative performance and customer basket size. The data indicates a distinct pattern that sheds light on optimizing sales strategies and understanding gender dynamics in sales.

The report emphasizes the significance of the relationship between sales representative productivity and customer basket size, as well as the effects of upselling and cross-selling tactics on performance. Notably, productivity increases steadily up to a threshold of about 70k as sales reps promote larger basket sizes. This emphasizes the importance of higher-value deals and implies that after a certain point, there is a point of diminishing returns. Customers’ resistance due to budget restrictions and sales representatives’ exhausted upselling efforts are two potential explanations. A more thorough examination of consumer behavior and post-70k sales strategies is required to understand the complexities of this plateau.

The analysis reveals complex order composition patterning. As the depth of purchased products rises to two, we withness a simultaneous increase in the productivity of our sales team - a correlation that aligns with our expectations. However, the subsequent revelation is rather astonishing. Beyond the two-products threshold, the productivity of our sales team appers to plateau, maintaining a relatively steady level up to the point of seven products. This peculiar pattern hints at a potential saturation point where the sales team’s eforts might be reaching optimal eficiency within this range of product depth.

While it’s understandable that increased product diversity could initially boost their performnce, this observed plateau might signify that other factors such as the complexity of managing a broader range of products could offset any further productivity gains. This findingd implies a crucial juncture for productivity improvement strategy.Rather than linear relationship between product depth and sales team performance, we might need to explore how factors beyond product variety influence their effectiveness.

Another important finding links the average conversion rates of the sales representatives to their overall productivity. We observed a direct correlation between the sales rep’s conversion rate and their productivity levels. When the conversion rate climbs, the average productivity of the sales rep follows suit, ascending in tandem. This findings aligns with our initial expectations - a salesperson’s proficiency in converting leads into customers undeniably contributes totheir overall productivity. Another revelatoin emerges as we venture further along the plot. Beyond around the 60% conversion rate mark, the rate of productivity improvement begins to diminish i.e the magnitude of productivity enhancement becomes less pronounced as conversion rates continues to rise beyond this point. These observation beckon us to reflect on the underlying dynamics at play.Could factors beyond conversion rate be having an exerting influence on overall productivity?

When you dig deeper into the analysis, a fascinating aspect becomes apparent: the comparison of sales rep productivity between genders. Contrary to popular belief, female sales representatives consistently outperform their male counterparts in terms of productivity when basket size is taken into account. This realization compels us to investigate the underlying forces causing this divergence. Possibile influencers include different communication philosophies, relationship-building skills, and alignment with client preferences. The solution to this intriguing gender-based productivity gap may involve specialized training programs designed to give male representatives access to the winning tactics that have made their female counterparts successful.

Furthermore, a closer look reveals that the gender-based performance gap is worse. Female sales representatives perform better than male counterparts across the board, which raises questions about subtleties in communication, building rapport with clients, and product knowledge. This distinction stands out more clearly as we examine data points with sales conversion rates of 40% and higher. In these situations, the difference between male and female sales representatives stands out more, suggesting that not only are female sales representatives skilled at maintaining high conversion rates, but they are also skilled at turning these successes into sizable gains in productivity.

In the realm of enhancing sales representative productivity, our rigorous analysis has unveiled pivotal insights that underscore the importance of both interaction frequency and duration. A significant correlation emerges between the frequency of customer interactions and the efficacy of our sales representatives. More monthly interactions correspond to higher productivity, highlighting the value of proactive engagement in driving tangible outcomes. However, a threshold of diminishing returns becomes evident beyond three interactions per month, emphasizing the need for a balanced approach.

Upon thorough examination of the relationship between contact time and average productivity, we identified a critical point where the relationship transitions. Initially, when the contact time inceases up to approximately 5mins, we discern a positive impact on the average productivity of our sales reps. this indicates that devoting more time to engage with potential clients during these intial moments leads to an upsurge in their performance. However, beyond the 5mins mark, the PDP illustrates a substantial decline in the average producivity of sales reps as contact time increase. This findings underscores the importance of optimizing the timing and duration of interactions with potential clients. While longer initial contact times can be beneficial to a certain extent, there appears to be an upper threshold beyond which extended engagement might not yield the desired results

When analyzing the time gap between sales calls made by our representatives to customer, a though-provoking trend emerges. Initially, as the time gap increases, we withness a slight but noticeable dip in the average productivity of our sales reps. This findings emphasizes the importance of maintaining a consistent engagement rhythm to sustain optimal productivity levels. However, when this time gap surpasses the 40-day mark. At this juncture, we start encountering a substantial and rather concerning decline in productivity. This reflection point underscores the critical threshold that exists beyond which the drop in productivity becomes more pronounced. This insights has profound implications for our sales strategy, indicating that prolong intervals between sales interactions could have detrimental effects on our team’s overall effectiveness.

Intriguingly, our comprehensive analysis of sales representative productivity delves into the nuanced realm of gender dynamics, revealing a layered perspective. The data highlights a striking equilibrium: both male and female representatives exhibit analogous levels of productivity when confronted with the same volume of customer interactions. This convergence underscores the absence of inherent gender-related differentials in the efficacy of customer engagement concerning productivity.

However, an enthralling revelation emerges when we scrutinize the interplay between interaction frequency and productivity across genders. Although initial interactions—up to two contacts per month—do not notably differentiate productivity levels between genders, a remarkable transformation unfolds beyond this threshold. Here, female representatives distinctly showcase elevated productivity compared to their male counterparts. This uncovers an avenue to harness the adeptness of female reps in scenarios necessitating amplified customer involvement.

This gender-based variation in productivity potentially emanates from diverse factors, encompassing communication styles, customer engagement strategies, and even the nature of the products in question. Delving deeper into these facets offers actionable insights into tailoring training, mentorship, and support to optimize the performance of both male and female sales representatives. Maximizing this revelation entails adeptly managing the frequency of customer interactions—refining timing, embracing data-driven segmentation, and adopting personalized approaches resonating with distinct customer demographics.

A deeper analysis reveals an even more nuanced panorama. Gender plays a pivotal role in delineating how the decline in productivity manifests. Notably, male sales representatives experience a more pronounced dip in productivity compared to their female counterparts as interaction durations exceed the optimal threshold. This discovery prompts pivotal inquiries concerning potential gender-specific communication styles, adaptability, and divergent approaches to extended interactions.

In light of these discernments, our project embarks on a multidimensional trajectory. Beyond discerning the optimal interaction duration for overarching productivity, it encompasses tailoring strategies that account for gender-specific dynamics to effectively counteract the observed downturn. Adapting training initiatives, communication methodologies, and introducing strategic breaks emerge as potential pathways to mitigate the productivity decline, particularly for male sales representatives.

Business Optimization

Scenario 1 In our pursuit of improving the productivity of the sales team, we have meticulously crafted a data-driven strategy that primarily concentrates on augmenting the efficiency of sales activities. By leveraging the insights gleaned from the Partial Dependence Plots (PDPs), we can effectively optimize various aspects of our sales process to achieve and sustain the desired outcomes.

Our overarching objective is to enhance the average productivity of sales representatives through adhering to the following parameters:

  • Contact Time: We aim to maintain a contact time between 5 to 10 minutes per interaction. This timeframe strikes a balance between engaging the customer effectively and maximizing efficiency.
  • Average Calls to Billing Ratio (AvgCalls_Bill): Our target is to uphold an average of 2 calls per billing. This helps in ensuring a sufficient level of engagement without overwhelming the customer.
  • Average Time btw Calls (Avgtime_Calls): We intend to sustain an average time span of 20 to 30 days for calls. This duration allows for meaningful follow-ups and relationship-building while avoiding excessive delays.
  • Visit Frequency: Our goal is to maintain a visit frequency of 2 to 3 interactions. This frequency enables us to remain on the customer’s radar without being intrusive.
  • Reach: We strive to achieve a minimum reach of 80%. This ensures that a substantial portion of our target audience is engaged, contributing to a more expansive market presence.

Implementation Strategy: To accomplish these goals and elevate sales representative productivity, we propose the following course of action:

Efficiency-focused Training: Provide comprehensive training sessions that emphasize efficient communication, active listening, and concise yet impactful engagement strategies. This will empower sales representatives to make the most of their contact time.

Smart Call Planning: Implement an intelligent call planning system that factors in customer preferences, history, and optimal timing for follow-ups. This approach will help in achieving the desired call frequency while maintaining relevance.

Data-driven Customer Segmentation: Utilize advanced data analytics to segment customers based on their needs, behaviors, and potential. Tailored engagement strategies for different segments will lead to more meaningful interactions.

Predictive Analytics for Reach Enhancement: Develop predictive models to identify the most promising leads and potential customers. By focusing efforts on these prospects, we can increase the likelihood of achieving the desired reach.

Feedback Loop Implementation: Establish a feedback loop where sales representatives can share insights and challenges they encounter during interactions. Regular refinement of strategies based on this feedback will contribute to continuous improvement.

Coordinate system already present. Adding new coordinate system, which will
replace the existing one.

Conclusion

In conclusion, our extensive analysis of sales representative productivity in relation to customer basket size has uncovered valuable insights that can significantly impact our sales strategy. The study has illuminated the intricate relationship between sales representative performance and customer basket size, underscoring the importance of optimizing sales strategies to capitalize on higher-value deals. The plateau in productivity beyond a threshold of around 70k in sales suggests the need for a deeper exploration of consumer behavior and post-threshold sales tactics. Additionally, the intriguing patterns in order composition and the interplay between conversion rates and productivity emphasize the complexity of factors beyond product variety and conversion rates that influence sales team effectiveness.

Furthermore, the gender dynamics within our sales team have unveiled surprising disparities in productivity. Contrary to prevailing assumptions, female sales representatives consistently outperform their male counterparts, raising questions about the communication and relationship-building skills that contribute to their success. The variations in productivity between genders at different interaction frequencies and durations present an opportunity to tailor training and support programs, leveraging the strengths of both male and female representatives. Notably, the observed decline in productivity beyond specific interaction thresholds suggests the need for nuanced approaches that consider gender-specific communication styles and adaptability.

As we move forward, our project’s trajectory will encompass optimizing interaction frequency, duration, and gender-specific strategies. By refining our understanding of these dynamics, we can develop more targeted training, communication methodologies, and strategic breaks to counteract the productivity decline and create a more balanced and effective sales team. Ultimately, these findings will play a pivotal role in reshaping our sales approach and fostering a more productive and inclusive sales environment.